Back to the Top
The following message was posted to: PharmPK
Dear all,
In a pharmacokinetic study using NONMEM, we have
elaborated a final regression model to estimate the
clearance of a drug A putting in consideration
different covariates.
We noticed that the inter-individual variability
(IIV)in drug clearance was about 23 % in final model
but in basic model it was about 17.5%. On the same
time, residual error for drug conc was less in final
model than basic model!.
Does anyone has explanation for the difference in IIV
between basic and final model? especially all methods
for validation of final model were promising?.
I suggest the use of sparse sampling strategy in our
study (trough concentration only was available)was the
reason for this difference.
Is there any other explanation?.
Thank you for help
Ehab EL Desoky
Pharmacology Dept
Faculty of Medicine
Assiut University
Assiut. Egypt
Back to the Top
The following message was posted to: PharmPK
Dr Ehab
How much was the difference in the residual variability? Also what model
have you used? All these will help in giving an explanation.
Venkatesh Atul Bhattaram
CDER, FDA.
Back to the Top
The following message was posted to: PharmPK
Ehab,
ehab eldesoky wrote:
> We noticed that the inter-individual variability
> (IIV)in drug clearance was about 23 % in final model
> but in basic model it was about 17.5%. On the same
> time, residual error for drug conc was less in final
> model than basic model!.
> Does anyone has explanation for the difference in IIV
> between basic and final model?
The key question I would have is why is the between subject variability
so low (with or without covariates)? Typical BSV in clearance in
patient populations is usually more like 50%. Were you studies done in
clones e.g. rats, medical students?
The increase in BSV when you added covariate(s) may be spurious.
NONMEM's estimates of BSV are known to be imprecise and a change of
this magnitude may be of no real importance.
Is there any good mechanistic reason to include the covariates in the
model or did you discover them by blindly testing everything you could
think of? If the latter then I would have no confidence in the meaning
of an improved objective function value.
> I suggest the use of sparse sampling strategy in our
> study (trough concentration only was available)was the
> reason for this difference.
I agree that your results are probably artifacts of the poor sampling
design. That depends a lot on the properties of the drug. It's probably
not a problem when the half life is long in relation to the dosing
interval (e.g digoxin) but can be problematic if there is substantial
change in conc during the dosing interval.
Dont believe it's good just because an FDA guidance mentions it.
http://www.fda.gov/cder/guidance/1852fnl.pdf
Sparse trough sampling strategies can be the worst possible way to
discover covariate effects on clearance. This study came from an FDA
pharmacometrician:
Lee PID. Design and power of a population pharmacokinetic study.
Pharmaceutical Research 2001;18(1):75-82
--
Nick Holford, Dept Pharmacology & Clinical Pharmacology
University of Auckland, 85 Park Rd, Private Bag 92019, Auckland, New
Zealand
email:n.holford.-a-.auckland.ac.nz tel:+64(9)373-7599x86730 fax:373-7556
http://www.health.auckland.ac.nz/pharmacology/staff/nholford/
Back to the Top
The following message was posted to: PharmPK
Dr Atul,
For clearance estimation, a proportional error model
was used.
In case of residual error, an additive model was
applied. The residual error values were as follows:
basic model: 6.8 mg/l
final model:5.2 mg/l
Dr. Ehab
PharmPK Discussion List Archive Index page
Copyright 1995-2010 David W. A. Bourne (david@boomer.org)